Research map / Video / Video generation
Video Object Segmentation-Aware Audio Generation
Ilpo Viertola, Vladimir Iashin, Esa Rahtu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add segmentation-aware control module to MMAudio | MMAudio · Segmented Music Solos (test, multi-source) | DeSync: 0.95 -> 0.44 | Improved |
| add segmentation-aware control module to MMAudio | MMAudio · Segmented Music Solos (test, multi-source) | IB-score: 35.94 -> 39.62 | Improved |
| add segmentation-aware control module to MMAudio | MMAudio · Segmented Music Solos (test, multi-source) | FD_PaSST: 530.60 -> 364.07 | Improved |
| add segmentation-aware control module to MMAudio | MMAudio · Segmented Music Solos (test, multi-source) | FD_VGG: 13.26 -> 17.19 | Hurt |
| add LoRA fine-tuning of DiT layers tied to segmentation features | SAGANet · Segmented Music Solos (test, multi-source) | DeSync: 0.44 -> 0.30 | Improved |
| add LoRA fine-tuning of DiT layers tied to segmentation features | SAGANet · Segmented Music Solos (test, multi-source) | IS: 3.34 -> 3.27 | Hurt |
| add LoRA fine-tuning of DiT layers tied to segmentation features | SAGANet · Segmented Music Solos (test, multi-source) | FD_VGG: 17.19 -> 19.36 | Hurt |
| add gated cross-attention instead of regular cross-attention (ablation) | SAGANet + LoRA · Segmented Music Solos (test, multi-source) | DeSync: 1.04 -> 0.30 | Improved |
| add gated cross-attention instead of regular cross-attention (ablation) | SAGANet + LoRA · Segmented Music Solos (test, multi-source) | FD_PaSST: 396.05 -> 330.01 | Improved |
| add segmentation mask to global video input (V -> V+M) (ablation) | SAGANet + LoRA · Segmented Music Solos (test, multi-source) | DeSync: 0.94 -> 0.96 | Hurt |
| add mask to global video input (V -> V+M), no LoRA (ablation) | SAGANet without LoRA · Segmented Music Solos (test, multi-source) | DeSync: 0.95 -> 0.56 | Improved |
| use only local focal crop instead of global video only (V -> V') (ablation) | SAGANet + LoRA · Segmented Music Solos (test, multi-source) | DeSync: 0.94 -> 0.32 | Improved |
Closest papers by meaning
- Sounding that Object: Interactive Object-Aware Image to Audio Generation (ICML 2025 · similarity 0.87)
- Customized Condition Controllable Generation for Video Soundtrack (CVPR 2025 · similarity 0.87)
- Read, Watch and Scream! Sound Generation from Text and Video (AAAI 2025 · similarity 0.87)
- V2Meow: Meowing to the Visual Beat via Video-to-Music Generation (AAAI 2024 · similarity 0.87)
- Tri-Ergon: Fine-Grained Video-to-Audio Generation with Multi-Modal Conditions and LUFS Control (AAAI 2025 · similarity 0.87)
- Improving Audio-Visual Segmentation with Bidirectional Generation (AAAI 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
